difference between `x in list` and `list.index(x)` for instances of anew-style class

[Riccardo Murri]

Hello,

I have some code that stops when trying to find a graph in a list of
similar graphs::

(Pydb) list
110 try:
111 canonical = self.base[self.base.index (graph)]
112 except ValueError:
113 raise ValueError, \
114 "Cannot find canonical representative for graph `%s`." \
115 - % (repr(graph),)
116
....

The list `self.base` contains "canonical" forms of the graphs and the
`graph` object must compare equal to some item of the list, which
indeed it does::

(Pydb) p graph == self.base[27]
True

(Pydb) p graph in self.base
True

However, I cannot directly get the index of the canonical graph (the
number "27" above was found by manual inspection)::

(Pydb) self.base.index (graph)
*** ValueError: list.index(x): x not in list

All graphs are instances of a `Graph` new-style class that implements
comparison operators `__eq__` and `__ne__`, but no other rich-compare
stuff.

I'm using Python 2.5::

Python 2.5 (release25-maint, Dec 9 2006, 16:17:58)
[GCC 4.1.2 20061115 (prerelease) (Debian 4.1.1-20)] on linux2

So my question is: what are the implementation differences between `x
in list` and `list.index(x)` and why can one report that an item is in
the list while the other cannot find its index? Should I add
something to the `Graph` class so that `index` works?

Thanks for any hint!
--
Riccardo Murri, via Galeazzo Alessi 61, 00176 Roma

[bpgbaires]

On 28 dic, 20:12, Riccardo Murri <riccardo.mu... @gmail.comwrote :

The list `self.base` contains "canonical" forms of the graphs and the
`graph` object must compare equal to some item of the list, which
indeed it does::

* (Pydb) p graph == self.base[27] *
* True

* (Pydb) p graph in self.base
* True

However, I cannot directly get the index of the canonical graph (the
number "27" above was found by manual inspection)::

* (Pydb) self.base.index (graph)
* *** ValueError: list.index(x): x not in list

All graphs are instances of a `Graph` new-style class that implements
comparison operators `__eq__` and `__ne__`, but no other rich-compare
stuff.

I'm using Python 2.5::

* Python 2.5 (release25-maint, Dec *9 2006, 16:17:58)
* [GCC 4.1.2 20061115 (prerelease) (Debian 4.1.1-20)] on linux2

So my question is: what are the implementation differences between `x
in list` and `list.index(x)` and why can one report that an item is in
the list while the other cannot find its index? *Should I add
something to the `Graph` class so that `index` works?

(I've checked on 2.5.1 but I don't see any relevant differences with
the 2.5 version). Looking at the source for both methods, they only
use the __eq__ operator, but there is a slight difference: while one
evaluates list[i]==x, the other reverses the operands. If your __eq__
is not reflexive, that could explain the difference.

class Graph(object):
def __init__(self, *items):
self.items = items

def __eq__(self, other):
if len(self.items) >len(other.item s): return False
return self.items == other.items[:len(self.items )]

pyList = [Graph(1,2,3), Graph(4,5,6), Graph(1,2,3,4)]
pyg = Graph(1,2)
pyg in List
True
pyList.index(g)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ValueError: list.index(x): x not in list
pyList[0]==g
False
pyg==List[0]
True

In your example, see if self.base[27]==graph is still True.

--
Gabriel Genellina

[Riccardo Murri]

bp*******@gmail .com writes:

On 28 dic, 20:12, Riccardo Murri <riccardo.mu... @gmail.comwrote :

>The list `self.base` contains "canonical" forms of the graphs and the
`graph` object must compare equal to some item of the list, which
indeed it does::

Â* (Pydb) p graph == self.base[27] Â*
Â* True

Â* (Pydb) p graph in self.base
Â* True

However, I cannot directly get the index of the canonical graph (the
number "27" above was found by manual inspection)::

Â* (Pydb) self.base.index (graph)
Â* *** ValueError: list.index(x): x not in list

All graphs are instances of a `Graph` new-style class that implements
comparison operators `__eq__` and `__ne__`, but no other rich-compare
stuff.

I'm using Python 2.5::

Â* Python 2.5 (release25-maint, Dec Â*9 2006, 16:17:58)
Â* [GCC 4.1.2 20061115 (prerelease) (Debian 4.1.1-20)] on linux2

So my question is: what are the implementation differences between `x
in list` and `list.index(x)` and why can one report that an item is in
the list while the other cannot find its index? Â*Should I add
something to the `Graph` class so that `index` works?

(I've checked on 2.5.1 but I don't see any relevant differences with
the 2.5 version). Looking at the source for both methods, they only
use the __eq__ operator, but there is a slight difference: while one
evaluates list[i]==x, the other reverses the operands. If your __eq__
is not reflexive, that could explain the difference.

That was indeed the reason: a bug in Graph.__eq__ broke reflexivity in
certain cases.

Thank you very much!!

--
Riccardo Murri, via Galeazzo Alessi 61, 00176 Roma

[Gabriel Genellina]

On 30 dic, 06:24, Riccardo Murri <riccardo.mu... @gmail.comwrote :

bpgbai...@gmail .com writes:

* (Pydb) p graph == self.base[27] *
* True
* (Pydb) p graph in self.base
* True
* (Pydb) self.base.index (graph)
* *** ValueError: list.index(x): x not in list

Looking at the source for both methods, they only
use the __eq__ operator, but there is a slight difference: while one
evaluates list[i]==x, the other reverses the operands. If your __eq__
is not reflexive, that could explain the difference.

That was indeed the reason: a bug in Graph.__eq__ broke reflexivity in
certain cases.

Combined with some other WTF bugs I've found at work, lately I feel
more like a detective than a software developer :)

--
Gabriel Genellina